Worms in the Cloud: How nematodes and AI are transforming R&D
Director, Biopharmaceutical Industry Solutions
NemaLife is using AI and Cloud to drive faster, lower cost biopharmaceutical R&D
Why Google Cloud?
Get unmatched cloud technology built on Google’s infrastructure.Learn more
The modest nematode Caenorhabditis elegans is a little worm with big potential.
Known as C. elegans, it was the first multicellular organism to have its full genome sequenced and has been an important contributor to the biological sciences since the 1960s (six people have won the Nobel Prize based on their work with the worm). C. elegans is considered a “model organism,” a non-human animal that is used to study biological processes, with findings used to provide insights into the functionality of other organisms. A millimeter in length and transparent, C. elegans shares about 40% of its genome with human beings.
In the hands of Lubbock, Texas-based startup NemaLife—and with a little help from artificial intelligence and the cloud—C. elegans may be able to transform the research and development process of a dozen industries including the food ingredients, cosmetics, and pharmaceutical sectors.
“We’re a techbio company that is developing platforms with microfluidics and AI,” said Marton Toth, chief business officer of NemaLife. “People have been imaging worms for the last couple of decades. Our goal is to extract all of this biological information and put it to use.”
Worms on a Chip
The basis of NemaLife’s technology is a microfluidic chip that provides the ecosystem for the nematodes. Microfluidics studies the behavior of fluids through micro-channels, and the technology of making the tiny devices through which flow small amounts of fluid (such as microliters). About 70 C. elegans can live on a NemaLife microfluidics chip, with a lifespan of two to three weeks.
The microfluidics chip is a departure from how nematode research has been traditionally performed. Historically, study of C. elegans or other nematodes has been done on agar plates, Petri dishes with a gel-like growth medium. The benefits of NemaLife’s microfluidic platform is that it allows for researchers to record and analyze the C. elegans behavior during the experiment, such as what happens when they are being fed test compounds.
“The worm is a millimeter in size so it’s amenable to these microfluidic chips,” said Dhaval Patel, director of research and innovation at NemaLife. “We can grow it in really large numbers. So we can deal with statistical populations that you’d never be able to deal with in mice. A single worm typically has several hundred progeny. We can run hundreds of chips a day, and we can do hundreds of animals per treatment. Those kinds of advantages don’t exist in any other kind of system.”
By working with nematodes for in vivo screening (testing with live animals, as opposed to in vitro which works with cellular or tissue samples), NemaLife can greatly accelerate the pace of initial research and development of new compounds. Whole-life experiments can be performed in two or three weeks, as opposed to months or years for animal testing using mice or rats. For a company that wants to test a lot of different compounds, C. elegans on a microfluidic chip offers an option to get more actionable results in less time.
Applying AI and the cloud to nematode research
Working with nematodes on agar plates used to be a very time consuming and manual process. Researchers would have to move the worms on the plate with instruments and enter measurements into spreadsheets by hand.
“From a historical context when people work with worms, they are looking at worms on an agar plate, down a microscope and taking physical records on a piece of paper,” said Patel. “And you have to have that person go in and copy all the data from that day’s work into a piece of software. That just takes so much time. I did experiments with thousands of worms for lifespan studies and just putting the data into the software took a full day after the experiment was done.”
This is where artificial intelligence and the cloud supercharge NemaLife’s process. Instead of manual observation, the worms can be continuously monitored via video or regular pictures throughout the experiment. And yet, the benefits of digital observation would go for naught if a human then had to manually go through and record all the worm’s behavior. NemaLife solves this by collecting the video data and processing it through artificial intelligence, cleaning it up so that it can be uploaded to a client dashboard.
“With this microfluidic engine, what we’ve attempted to do is build a data pipeline,” said Siddhartha Gupta, lead software engineer at NemaLife in charge of building the company’s data and AI solutions. “What that allows us to do is minimize costs and organize the data. We focus on the ontology of the data, we focus on the accessibility. With that in mind we moved most of our compute to the cloud.”
When we have things in the cloud, one of the advantages of the centralization is that we can train models on it.
Siddhartha Gupta, Lead Software Engineer, NemaLife
NemaLife uses Colaboratory, a data science tool from Google Research that allows anybody to write and execute python code through the browser. It is especially well suited to machine learning and data analysis. Once the data has been wrangled, NemaLife can use a machine learning model called ResNet-50, a pre-trained convolutional neural network that is 50 layers deep.
“ResNet-50 improved our bottom line a lot,” Gupta said. “We are able to offer much better detection and our costs are lower. We can now build on top of that to have more behavioral metrics. We have this end-to-end pipeline where we have this ingestion, transformation and loading of the data for clients that is all done in the cloud now.”
Expanding beyond nematodes
Based on the powerful combination of its microfluidics platform and AI and cloud tools, the team is confident that it can expand both its product line and the industries it serves as it continues to grow.
“We are trying to build an operational pipeline for machine learning where the models are automatically trained and automatically updated,” said Gupta. “That allows us to diversify what we can test with this microfluidic engine. Tomorrow if we used bubbles instead of worms, we could probably train a model to analyze the fluid dynamics of that.”
In addition to in vivo screening, NemaLife is confident it can tackle in vitro screening as well as physical characterization work to help accelerate companies’ research and development programs.
“It’s not always the same chip,” said Patel. “The expertise allows us to design custom chips to meet custom needs.”
NemaLife recently announced a strategic investment from Boston-based Motif FoodWorks to use the microfluidics platform to test ingredients for plant-based foods. The food and nutraceutical market is fertile ground for NemaLife, but the microfluidics platform—with or without the nematodes—will allow the company to expand to a dozen or more industries.
“When you think about getting information to allow you to make decisions faster, that is a competitive advantage,” said Grant Campany, NemaLife’s chief financial officer. “If we can do in a few days what might typically take months, there is a tremendous strategic opportunity for these companies.”
For NemaLife, the goal is to fundamentally change how research and development is done, accelerating the pace of experiments while cutting costs.
“We want to take this technology to the level where we can very significantly transform the whole R&D industry,” said Toth. “Whether it’s in-vivo, in-vitro, or physical characterization. The savings advantage that we’re talking about, it’s not 10 or 20 percent. It’s orders of magnitude.”